Twenty years of research on shrinking cities: a focus on keywords and authors

Abstract

Most scholarly attention to vanishing cities is fairly recent so, to guide future research, a comprehensive evaluation of prior findings is required. This study is a network analysis of 333 publications authored in English, published over the last two decades. The findings are as follows: (1) shrinking city research has increased significantly since 2016; (2) the key themes are planning, decline, depopulation, policy, regeneration, vacant land, green infrastructure, and case studies such as Detroit; and (3) major academic groups have not yet collaborated effectively on the subject.

Introduction

Defining shrinkage

Numerous cities are massively depopulating or shrinking (Hartt, Citation2019). The term ‘shrinkage’ began appearing more often in the 1980s after a period when several U.S. cities and counties had begun to lose residents, mainly due to deindustrialisation (especially from 1950 to 1980). Around the same time, German scholars applied the concept of urban shrinkage following the fall of the Berlin wall and unification in 1990 (Alves, Barreira, Guimarães, & Panagopoulos, Citation2016) and the collapse of the Soviet Union and the shrinkage of post-socialist cities in Eastern Europe (Schett, Citation2011). Such societal changes have prompted widespread observation of the spatial and temporal patterns, causes, consequences, and responses to urban shrinkage.

The neighbourhood life cycle theory, developed in the early 1920s by urban sociologists, explains urban decline as an inevitable part of a city’s life cycle (Lang, Citation2000). Urban neighbourhoods worldwide experience population growth, stabilisation, and decline, making depopulation a natural phase of urbanisation. According to the theory, long-established cities such as London, Tokyo, and Paris have communities that necessarily declined or depopulated to make room for new growth (in their case, in the 19th century). On the contrary, continuous decline theory considers the disproportionate developments between territories (e.g. suburbanisation and edge cities), economic transitions (e.g. the relocation and loss of manufacturing jobs), or demographic transitions (e.g. ageing and low fertility) as significant drivers of decline (Alves et al., Citation2016), which differs significantly from the degree of natural decline. Both theories attempt to justify the process of urban shrinkage, but the latter is better suited to the multifaceted nature of recent urban decline due to its different causes, consequences, and dynamics compared to precedents.

Many cities have risen and fallen throughout history, but many of them also recovered. Recently emerging shrinking cities, however, may be incapable of escaping decline in the short term due to the complexity of interrelated underlying problems including more unpredictable societal changes, globalisation, a long-established industrial history, and related skills or lack of broad skill ranges among its residents (Rybczynski & Linneman, Citation1999). Traditional economic and policy forces, mainly concerning economic development and physical redevelopment, have not always succeeded in reversing trends of decline. For example, the relocation of Quicken Loans and the massive investment from JP Morgan’s Corporate Social Responsibility program, while partially successful, has not fully reversed the shrinkage-induced decline in Detroit (Koven & Koven, Citation2018). Such efforts can sometimes actually backfire, further depleting the city’s resources (although not in the case above), and eventually harming the quality of life among remaining residents (Park & LaFrombois, Citation2019), and those revitalisation efforts can actually lead to more inequality (LaFrombois & Park, Citation2023). Therefore, some planners and scholars are now promoting an urban paradigm shift towards ‘smart decline,’ also known as smart shrinkage, right-sizing, or city diet (Park, Kim, & Seong, Citation2021). Popper and Popper (Citation2002), who first coined the term, ‘smart decline,’ suggested applying urban rescaling strategies towards something smaller or right-sizing. Although decline-oriented strategies are still new for many cities, several depopulating cities in the U.S. started to adopt some right-sizing tools (LaFrombois, Park, & Yurcaba, Citation2019). Adopting these tools signals that cities are changing their orientation towards a more agile approach to decline through lot management and temporary reuse to avoid future downturns.

The growing body of knowledge about shrinkage

Research on shrinking cities has rapidly grown recently, with an increasing focus on policy as a primary intervention strategy. There are several comprehensive reviews of the existing research focussing on shrinking cities, which has studied the symptoms, causes, and solutions in various countries. Mallach (Citation2017) observed the trajectories and policy responses of shrinking cities in the U.S. after World War II, while Berglund (Citation2020) explored Japan, Germany, and Rustbelt cities and called for the expansion of neo-liberalization under austerity. In particular, he highlighted three major practices: planned shrinkage, city greening, and vacant property reuse, which arose in opposition to pro-growth strategies.

Several special issues have also been published on shrinking cities in various journals, representing the evolving importance of the topic in urban studies: ‘Shrinking Cities and Towns: Challenge and Responses (2013, Urban Design International),’ ‘Shrinking Cities: Rethinking Landscape in Depopulating Urban Contexts (2017, Landscape Research),’ ‘Shrinking Cities from Marginal to Mainstream: Views from North America and Europe (2018, Cities),’ and ‘Promoting Social Justice and Equity in Shrinking Cities (2020, Journal of Urban Affairs).’ These special issues commonly considered shrinking cities as a global phenomenon and emphasised the necessity of academic discussion on the topic across different fields. In particular, Lima and Eischeid (Citation2017) edited a special issue of Landscape Research, introducing seven articles and one photographic essay addressing innovative approaches to the balanced management of population and resources to explain the role of landscape as a concept and landscape architecture as a discipline. They noted that the issue papers focus mainly on sites left vacant by urban shrinkage. For example, after analysing the distribution of non-productive spaces in the sprawling city of Fort Worth in the US from 1990 to 2010, Newman and Kim (Citation2017) emphasised the possibility of repurposing those abandoned spaces for ecological and cultural benefits, even though those spaces seemed to have a low potential for redevelopment, at least by traditional economic criteria. Albro, Burkholder, and Koonce (Citation2017) also explored small-scale and dispersed vacant spaces in Rust Belt cities in the US and suggests a way to use those spaces as stormwater management assets.

In contrast to these case studies or special issues, Döringer, Uchiyama, Penker, and Kohsaka (Citation2020) attempted a more comprehensive and systematic literature review. They explored 100 case studies in 70 articles (2005–2017) in the EU and Japan to investigate urban shrinkage and its causes, outcomes, and responses using qualitative and comparative meta-analyses. They found that early studies focussed more on the causes and effects of shrinkage, whereas later research highlighted governmental policies and reactions to the phenomenon. Deindustrialisation, out-migration, and post-socialist transformation were the most influential driving factors in European cases, while low birth rate, suburbanisation, and ageing were the most critical in Japanese cases. It was found that urban shrinkage led some European countries to suffer from increases in unemployment, housing vacancies, and economic decline, while urban decay (e.g. run-down buildings) and housing vacancies resulted in Japan. Overall, the geographical differences found in urban shrinkage encouraged cross-national comparisons and international cooperation on the topic.

Most reviews on urban shrinkage concentrated on essential issues in shrinking cities, such as definitions, symptoms, causes, impacts, and solutions. Studies, however, have yet to examine trends systematically and quantitatively in academic studies in shrinking cities. Some ideas may have been missed or failed to discover supporting evidence because earlier qualitative reviews relied primarily on the level of researchers’ insights or knowledge in relevant issues (perspective-based techniques). For example, Luescher and Shetty (Citation2013, p. 1) said, ‘a wider population is engaging in this, and discussions of the future of these communities are slowly entering the public discourse through diverse channels.’ It is valuable guidance, but readers may want to hear more what about ‘a wider population,’ ‘these communities,’ or ‘discourse’ mean, for instance, which researchers are currently collaborating on which topics, and what kinds of discourse have been shared.

As such, a quantitative approach through a network analysis (NA) seeks to fill this gap while simultaneously constructing, visualising, and investigating the networked knowledge on the primary topics discussed and the academic communities studying shrinking and declining cities. Applying NA to keywords and author lists from 333 papers, written in English, and published from 2001 to 2020, this study explores trends in shrinking city studies over two decades to answer two questions: (1) what primary keywords characterise the field and/or require further examination and (2) which academic communities have been most active and/or how can they effectively work together to respond to challenges related to the global issue of urban decline? The findings will offer a better understanding of the present state of the research agenda on declining cities (over the previous 20 years) and facilitate further discussion of issues that remain unsolved. Findings from the study are also intended to bridge between international research groups to advance networking, fostering increased future research and practice in various urban studies fields such as policy, planning, and landscape architecture.

Research methods

Data collection

Google Scholar, the Web of Science (WoS), and Scopus are all recognised by researchers as the largest and most reliable databases for scholarship; WoS and Scopus are the most frequently used for meta-analyses (Zhu & Liu, Citation2020). As Martín-Martín, Orduna-Malea, Thelwall, and López-Cózar (Citation2018) found, although all three databases overlap to some extent, Google Scholar produces more non-journal-type resources such as books, conference proceedings, and unpublished works than the others. Scopus has articles dating back to 1960 and offers access to 12,850 journals, whereas WoS covers 8,700 journals (Falagas, Pitsouni, Malietzis, & Pappas, Citation2008), and its planning-related records are only from 2006 or later. Because the study considered a two-decade period from 2000, the Scopus database was used only to search for keywords from published academic journals written in English.

Initially, papers published from 2000 to 2020 with ‘shrinking city,’ ‘depopulating city,’ or ‘declining city’ in titles, abstracts, or keywords were searched. Using only these three comprehensive search strings avoided the large task of selecting keywords. Scopus searches for lemmatised (plural/adjectival) forms and ignores punctuation and stop words (e.g. the, it, of) (Reed College Library, Citation2021). Among returned 440 publications, 35 papers were deemed irrelevant and removed during the initial search, following manual review of titles and abstracts. Some of these were published in environmental journals and mentioned shrinking ecological regions, while others simply referenced a shrinking city as a study area to investigate pollinator population declines, rather than the shrinkage of the city itself. Two papers were also excluded for not being written in English. Seventy articles were then excluded because they lacked keywords. Finally, 333 published articles remained for analysis (Appendix 1 for the complete list).

The author keywords from some articles were normalised in the following ways: plural nouns were made singular, British English was made American, city names were simplified, state names were omitted, the word ‘urban’ was omitted in compound keywords, and synonymous phrases were rephrased. Such changes were kept to a minimum so as to not lose the researchers’ intentions. For instance, ‘vacant lots,’ ‘vacant properties,’ and ‘vacant land’ were kept as compounds because deleting ‘vacant’ would alter the meaning. For the complete list, two steps (Appendix 2). The above process resulted in 1,778 keywords for the quantitative analysis. Also, author names were collected without middle initials; 598 authors were found in total.

Analytical methods: burst detection & network analysis

Research trends were explored using burst detection (BD) and NA. BD can identify concentrated trends in research over a specific period using time-series data (Zhu & Shasha, 2003). Kleinberg (Citation2003) first constructed this algorithm, and other researchers (Sohrabi, Vanani, Jalali, & Abedin, Citation2019; Tamura, Tamura, Kitakami, & Hirahara, 2012) found that BD could effectively indicate the rise and fall of specific topics in academic articles over periods. This study uses Kleinberg’s algorithm (i.e. γ = 0.7 and density scaling = 3.0) to calculate each bursty word’s duration and weight; higher weights indicate relative frequency and longer duration (Kim, Khan, Wood, & Mahmood, Citation2016). Burst analysis helps to identify periods when the incidence of particular keywords was temporarily higher than usual.

Secondly, NA was done to identify clusters of keywords and authors composing networks. NA is widely used to disclose the structure of knowledge in a given field and the collaborative patterns among academics (Khan & Wood, Citation2015), thanks to its effectiveness for visualising and exploring the connections among large volumes of data (Park & Newman, Citation2017). A quasi-network matrix of keywords was constructed to create the network, which comprises links between keywords—represented by nodes—in the same article (Kho, Cho, & Cho, Citation2013). A total of 904 nodes and 3,738 links were created. In the same way, the author network was consisted of 598 nodes and 1,070 links. Each node represents a keyword or author, and links indicate co-occurrence of nodes. Centrality measures such as degree, eigenvector, betweenness, and modularity scores were observed to identify not only keywords and author clusters but also key persons between groups. Degree centrality scores aid in finding nodes with the most links, while eigenvector centrality identifies the importance of nodes in terms of their connectedness to other essential nodes (Golbeck, Citation2013). Betweenness centrality denotes a network’s bottlenecks of a network as in its connections among its different parts (Park, Citation2015). Modularity calculates the adequacy of a division of a network into modules: high modularity scores indicate dense connections between nested nodes within a module but fewer connections with nodes in other modules (M. E. Newman, Citation2006). For analysis and visualisation, Sci2 tool (Sci2 Team, Citation2009) and Gephi 0.9.2 (Bastian, Heymann, & Jacomy, Citation2009) were used. The author and keyword clusters identified were classified as either ‘problem-oriented’ or ‘location-oriented.’

Results

Research trends and topic burstiness

Until 2011, fewer than ten papers were published, per year, related to shrinking, declining, or depopulating cities. The rate began to increase dramatically in 2016 (Figure 1); after 2016, they entered what could be called the growing phase. Schilling and Logan (Citation2008)’s work, which mostly described green infrastructure in vacant lots, including land banks and collaborative neighbourhood plans for shrinking cities, had the most citations, with 352. Next, Wiechmann and Pallagst (Citation2012)’s study of the policies and planning strategies in Europe and the U.S., had 208 citations. During the growth phase, Haase, Bernt, Großmann, Mykhnenko, and Rink (Citation2016)’s study, which focussed on the drivers, downsides, and consequences of shrinkage in Europe, was cited 112 times, followed by Martinez-Fernandez et al. (Citation2016), which compared local policies across Australia, Japan, and Europe, with 65 citations.

Figure 1.

Total numbers of papers and keywords (2001 ~ 2020).

The dynamics of topic usage were also explored (Figure 2). Among the top-ranked 22 bursting keywords with the highest weights, the top five significant emerging and disappearing keywords were ‘Eastern Germany’ (2009–2013), ‘Governance’ (2013–2014), ‘Neoliberalism’ (2013–2015), ‘Deindustrialization’ (2009–2015), and ‘Gentrification’ (2009). ‘Deindustrialization’ appeared over the longest period, followed by ‘Eastern Germany,’ ‘Residential Mobility,’ and ‘Adaptation’ over five years during the initial phase, and several terms such as ‘Gentrification’ (2009), ‘Urban Theory’ (2016), and ‘Sustainability’ (2016) had short lifespans; overall, words that first appeared during the growth phase had relatively momentary lives.

Figure 2.

Top burst keywords with highest weight.

Co-occurrence of keywords

Table 1 presents the identified centrality scores, and Figure 3 illustrates connections among nodes using bubble diagrams and connected links. The size of the bubble, the distance between bubbles, and the thickness of the lines express the relative importance of each keyword within an entire network. Keywords with high eigenvector and degree scores were ‘Shrinking City,’ ‘Shrinkage,’ ‘Decline,’ ‘Planning,’ ‘Depopulation,’ ‘Regeneration,’ ‘Policy,’ ‘Vacant Land,’ ‘Green Space,’ and ‘Green Infrastructure.’ These keywords were likely to be positioned at the central parts of the whole network. Meanwhile, keywords such as ‘Educational Leaders,’ ‘Social Justice,’ ‘Urban Fabric,’ or ‘Shrinking Small Towns,’ showed low eigenvector and degree scores.

Table 1.

Top keywords with high frequency, Eigenvector Centrality Score, and cluster.

Frequency		Eigenvector		Degree
Shrinking City	201	Shrinking City	1.000	Shrinking City	568
Shrinkage	53	Shrinkage	0.314	Shrinkage	153
Decline	37	Decline	0.269	Decline	112
Planning	24	Planning	0.223	Planning	80
Depopulation	23	Depopulation	0.193	Depopulation	77
Vacant Land	18	Policy	0.185	Regeneration	62
Regeneration	17	Regeneration	0.181	Policy	62
Policy	16	Detroit	0.156	Detroit	60
Detroit	14	Vacant Land	0.153	Green Infrastructure	52
Demographic Change	13	Green Infrastructure	0.143	Vacant Land	51
Housing Vacancy	12	Japan	0.139	Green Space	47
Green Infrastructure	12	Green Space	0.134	Housing Vacancy	44
				Japan	44

Figure 3.

Network of keywords and clusters (Force Atlas Layout). *The size of the node circle corresponds to the degree centrality score. Only nodes with a degree centrality value of ten or greater are illustrated. **Colour coding of each node is used to denote nodes in the same cluster.

By comparing the field’s developmental phases in succession, we can see that the study areas and topics have changed. During the initial phase, ‘Policy,’ ‘Deindustrialization,’ ‘Detroit,’ ‘Planning,’ ‘Demolition,’ ‘Eastern Germany,’ and ‘Depopulation’ hold the highest eigenvector and degree scores. During the growing phase, ‘Regeneration,’ ‘Vacant Land,’ ‘Green Infrastructure,’ ‘Housing Vacancy,’ ‘Japan,’ and ‘China’ ranked in the top ten: ‘Depopulation’ remained, but ‘Demolition,’ ‘Eastern Germany,’ and ‘Detroit’ disappeared.

Several distinctive clusters were also found in the modularity class analysis (Figure 3). The five largest clusters were titled based on words that showed the highest degree score (Table 2). Note that the interpretation should be based on the general characteristics of the keyword clusters, rather than just focusing on the cluster names. The titles were only provided to identify the uniqueness of each cluster. ‘Regeneration & Policy’ holds the most keywords (30.4% of all nodes). This cluster contains topics that cover the overall reasons for (e.g. deindustrialisation, suburbanisation, or immigration), consequences of (e.g. vacant land, gentrification, or demographic change), and policy responses towards (e.g. regeneration, green infrastructure, demolition, or right-sizing) urban shrinkage. Detroit, as well as some other US cities (e.g. Buffalo, Cleveland, or New Orleans), Eastern Germany, and China were included in this cluster. In the second-largest cluster, ‘Governance’ (15.0%), urban politics, urban regime, or policy responses played a great role in post-socialist cities such as Ostrava in Czechia, Łódź in Poland, and cities in northeast China, held several nodes. The last three clusters showed similar sizes; these held approximately 6% of the network’s nodes. The ‘Housing Vacancy & Demography’ cluster followed. This cluster includes keywords such as ‘housing vacancy,’ ‘abandoned house,’ ‘household,’ and ‘population.’ These keywords were associated with Latin American (e.g. Mexico, Guadalajara-Mexico, or Brazil) and European (e.g. France, Glasgow, or Switzerland) cities and countries. ‘Greenspace’ emphasises parcel-scaled regeneration processes, particularly changing non-green land uses into green reuse. Lastly, ‘Ecosystem’ highlighted the potential of underused vacant land in shrinking cities for biodiversity, habitat management, and urban wilderness. The keywords pertained to research on ecological and sociological phenomena that occur naturally in abandoned areas, as well as to countermeasures for vacant land and their potential impact on ecosystems, such as the compositions of bee communities.

Table 2.

Top clusters, keywords, and study area.

Cluster	Regeneration & policy	Planning& governance	Housing vacancy & demography	Green space	Ecosystem
Top Study area	Detroit & Eastern Germany	Post-socialist City	Europe and South America	Finland	Post-Soviet country
Top keywords	Shrinking city	Shrinkage	Housing vacancy	Green space	Brownfield
	Decline	Planning	Europe	Quality of life	Post-Soviet country
	Depopulation	Governance	Household	Ageing	Sustainability
	Policy	Growth	Population	Residential mobility	Vacant lot
	Regeneration	Post-socialist city	Greening	Housing market	Post-industrial
	Detroit	Inner city	Out-migration	Invasive species	Lead
	Vacant land	Ostrava	Stadtumbau*	Housing stock	Bee Conservation
	Green infrastructure	Urban regime	Australia	Building stock	Pollinator
	Urbanisation	Cross-continental comparative perspective	Job creation	Habitat management	Biodiversity
	Demolition	Revitalisation strategy	North America	City	Energy performance of buildings
	Ecosystem services	Path dependence	Mexico	Social capital	Urban
	Demographic change	Ageing population	Abandoned house	Environmental affordances	Arthropods
	Eastern Germany	Urban politics	Urban forestry	Green open space	Mowing
	Historic preservation	Planning strategy	NDVI	Living environment	Null model
	Gentrification	Spatial structure	Urban and regional development	Retirement migration	Pocket Prairies
			Sustainable land use		Predators

^*Urban Regeneration East program in Germany that mainly supported the demolition of vacant housing (Bernt, Citation2019).

Author clusters

In all, 598 authors are connected by 1,068 edges, but the network was sparsely connected. Clusters A–D formed the most significant and large sub-network among nine clusters with a modularity score over 2.0% (e.g. clusters A–I in Figure 4). Clusters A–D are strongly connected with each other, while the rest formed relatively independent clusters: clusters F, G, and I comprise Chinese researchers; clusters E and H comprise researchers from the U.S. From the location of the active researchers’ institutions and each cluster’s research theme, clusters were roughly named to indicate their overall characteristics (Figure 4). The research topics and location of affiliations within a cluster are not, however, limited to those in their titles.

Figure 4.

Author clusters and network (Fruchterman Reingold Layout).

As shown in Table 3, researchers with higher betweenness scores, such as Pallagst K., Wolff M., Wiechmann T., London J., and Schwarz K., played a vital role in connecting clusters A–D into one giant community. Hasse A. and Hollander J., Pallagst K., and others of high degree in the network exhibited the most connections in the network, indicating that they have worked actively with other researchers. Some of them, such as Pallagst K., Hollander J., and London J., also exhibited high eigenvector centrality score, which implies that they could influence the whole network significantly. Combined centrality scores suggested that Pallagst K., Hollander J., and London J. are essential actors in shrinking-city studies. The earliest of such studies were led by European researchers such as Hasse D. and Rink D. in cluster D, and Rousseau M. and Fol S. in cluster C; more recently, the members of clusters G–I have added several shrinking city studies.

Table 3.

Top ten authors with High Centrality Scores.

	Degree		Eigenvector		Betweenness
Haase A.	22	Pallagst K.	1	Pallagst K.	4227.5
Hollander J.	21	Hollander J.	0.988	Wolff M.	4180.91
Pallagst K.	21	London J.	0.979	Wiechmann T.	4000.667
Rink D.; Wolff M.; Gardiner M.; Haase D.	19	Watson V.; Blanco H.;Alberti M.; Olshansky R.; Chang S.; Wheeler S.; Randolph J. ; Schwarz T.; Popper F.; Parnell S.; Pieterse E.	0.93	London J.	2392
			"2157.000"
			"1443.681"
			"1214.476"
Allen C.; Garmestani A.	18		"Hollander J."		"1091.000"
	"Rumpel P."		"913.500"
London J.	16		"Fol S."		"729.667"

Discussion

This study observed the endeavours of academics in shrinking city studies by systematically investigating the themes, which can be inferred from keywords, and discovering the active researchers. Through analysis of the incidence frequencies, bursts and network relations among keywords and authors, this paper could help illuminate trends in shrinking cities studies. This paper, however, should be read cautiously, as it covers only papers written in English, via a limited range of identifiers, and the number of related studies has increased considerably after the observation period. Further, major works in social science are frequently published as books, and shrinking cities are no exception.

Trending topics in shrinking cities and future direction

Following the characteristics of each keyword, topic clusters from the research on shrinking, depopulating, and declining cities can be considered as ‘problem-oriented’ or ‘location-oriented.’ If interpreting clusters as ‘problem-oriented,’ clusters can be divided into five large clusters: (1) policy and regeneration efforts in shrinking cities, (2) planning strategies and governance, (3) housing vacancy and demography, (4) greening, and (5) ecology. Most of the top-ranked keywords were also the main nodes of each cluster. The ‘Regeneration and Policy’ cluster, for example, embraces various topics, but two main sub-subjects, narratives of shrinking cities policy responses (e.g. regeneration, historic preservation, demolition, green infrastructure, or right-sizing) were frequently presented. Papers with those keywords blamed the endeavours of urban regeneration based on growth-oriented planning paradigms like megastructure-led planning (Slach, Nováček, Bosák, & Krtička, Citation2020), centralised planning (Couch & Cocks, Citation2013), and middle-class-focussed planning (Mah, Citation2021). Many urged an abandonment of these paradigms, via a reconceptualization of decline and alternative response (Blanco et al., Citation2009), which calls for the needs of a paradigm shift towards low-growth trajectories while increasing governmental intervention changes from growth coalitions (i.e. driven by the market and private companies) to grant coalitions (i.e. state-level supports and spending led) in areas with both weak markets and actors (Rink et al., Citation2014). Landbanking in the U.S. is an excellent example of how the government could exercise great power to fight for urban vacancy and how the state-level land bank statute could effectively mediate urban issues. Bottom-up efforts such as community participation and neighbourhood-based non-profits and outreach efforts for regeneration (Kim, Newman, & Jiang, Citation2020), local grassroots movements, and community empowerment (Agirre-Maskariano, Citation2019) were also mentioned. Housing vacancy—the most visible consequence of decline—was closely related to greening and the ecosystem. Before the official planning process is activated, several local efforts are made to investigate prospects for abandoned areas, with creative means of decreasing waste and re-greening being a significant focus. Many see the proliferation of vegetation after demolition as a new opportunity for ecological (e.g. habitat provision, stormwater runoff reduction, and atmospheric carbon storage) and sociological benefits (e.g. well-being, recreation, and interactions) (Haase, Citation2008; Riley, Perry, Ard, & Gardiner, Citation2018), and more permanent uses like urban agriculture were frequently suggested. As Morckel (Citation2017) proposed, sustainability analysis would guide the way to select and prioritise the areas for naturalisation. In future analysis, additional evaluation variables could be offered from the perspective of landscape architecture. The investigation by Hino, Yamazaki, Iida, Harada, and Yokohari (Citation2023) on the relationship between urban landscapes and physical activity in shrinking cities serves as a starting point for future research on the impact of the landscape on various aspects of well-being, including the creation of new employment opportunities, the increase in social capital, the promotion of co-sharing, and the improvement of governance in the greening of vacant and abandoned properties.

Moreover, topics that have seen recent bursts and are located at the periphery of the network should not be treated as unimportant: their growth trends should be taken into account. Such topics may be diverse, ranging from greening (e.g. urban ecology, community garden, and green infrastructure) to depopulation (e.g. spatial and temporal change) and vacancy (e.g. projection, spatial pattern, and drivers), inequality, ethnicity, segregation, climate challenge, hazard, or governance, and could be related to environmental and socio-economic disparities.

If deducing clusters as ‘location-oriented,’ the study areas related to the clusters were mainly limited to Germany, the U.K., and the U.S. This may be a reasonable set of focus considering the economic decline of the 1990s and 2000s. East Germany experienced emigration to West Germany before reunification in 1990 (Rieniets, Citation2009). Then, the collapse of the state-directed economy resulting in deindustrialisation and high unemployment, induced substantial adverse shocks (Rink et al., Citation2014). Post-socialist cities in Eastern Europe followed paths similar to that of East Germany, and the drivers and consequences of shrinkage were examined in the early stages (Popescu, Citation2014; Wiechmann & Pallagst, Citation2012). On the other hand, the fall of old industrial towns and legacy cities, such as Detroit in the U.S., began to enter urban discourse in the late 1990s, even though the phenomenon actually began after WWII (Mallach, Citation2017). Meanwhile, few works have examined Russia, South Korea, Iran, Australia, or Canada. Li and Mykhnenko (Citation2018) described the unique role of the state in, for instance, industrial restructuring (from coal, steel, or petrochemical to high-tech manufacturing), the core–periphery development gap, and intended shrinkage through planned resettlement for urban regeneration or megaprojects like the construction of hydroelectric dams in China. Their study suggests various strategies for addressing urban decline in different governmental structures. As such, if past studies focussed on observing the phenomenon in and responses from different countries, more empirical and microscopic studies that examine the governmental interventions, practices, and policies across the different continents at the local, state, and federal levels should be done in future to test suggested remedies and identify practical barriers to putting right-sizing strategies into effect.

Current research network and future direction

The author network might be used to identify how academic groups in shrinking city studies have interacted. The largest clusters were found from U.S. and Germany, where the key players were Hasse A., Hollander J. and Pallagast K., Gardiner M., Allen C., and Garmestani A. In particular, Pallagst K. Wolff M. and Wiechmann T. connected researchers in different clusters. Researchers in the same cluster showed the similar research interests and locations. This may be due to the fact that individuals who attend the same school or work at the same place are likely to have better collaborations as this can result in more efficient and a smoother working relationship; for instance, Pallagst K., Hollander J. and Popper F. graduated or worked from the same school, Rutgers University, US.

The results of this study, however, suggest that project groups, coalitions, and collaborations can help expand the research network beyond research interests, regions, or academic communities. For example, the majority of leading researchers in Cluster D, which we labelled ‘Germany Ecology,’ are members of the SHRINK SMART, an EU research network actively working on urban shrinkage. Members Fol S. and Cunninham-Sabot E. of Cluster C, named ‘France Urban Policy,’ also belong to the SHRINK SMART network as advisory board members. This EU research network has brought together researchers from different countries, such as the project-leader group from the Helmholtz Centre for Environmental Research – UFZ in Germany, Couch C. and Cocks M. from Liverpool John Moores University in the UK, Rumpel P. and Slach O. from the University of Ostrava in the Czech Republic, Krzysztofik R. from the University of Silesia in Poland, Mykhnenko V. from the University of Nottingham in the UK, and Nadolu B. from the West University of Timisoara in Romania (Shrink Smart, Citation2023). Other EU research projects, such as the Shrinking Cities International Research Network (SCiRN), CIRES, and Re-City (Armstrong & Kapp, Citation2010), can also serve as good examples. For instance, Pallagst K. in Cluster B has a strong link with Wiechmann T. in Cluster C, as they were both members of SCiRN and Re-City. As Großmann, Bontje, Haase, and Mykhnenko (Citation2013) called for, a global discussion is necessary since the challenges of shrinking cities result from the global dynamics of demographic and economic changes, resulting globally in widespread phenomena. Therefore, relatively small groups (Clusters E–I), primarily from China, the United States, South Korea, and Hong Kong, need to connect with key players in other clusters to expand the nature of the field and to advance interdisciplinary research efforts.

As stated at the beginning of the Discussion, the results of keyword and author network analysis suggested in this study make it challenging to review all related studies comprehensively. However, the suggested method does already help to systematically review the essential areas of work, study trends, and collaboration patterns. If comparable research is conducted in ten years, it will be interesting to see how the network has evolved, how the subjects have changed, and how the primary players have changed.

Conclusion

The observation of keywords from scholarly studies for two decades exposes many topics on which future research should focus. First, many studies have been conducted on only a few cities (primarily legacy cities) in the U.S., Eastern Germany. Studies of cities in Asia or Oceania are expanding, but much more is needed. Expansion of research cases across countries is also necessary since each city may have unique mechanisms contributing to its decline, and particularities of its political background may result in different policy interventions between countries and municipalities. The cities included in existing studies are relatively large and are primarily aged industrial-based urban areas. Concern is growing around smaller and medium-sized cities. Many of them are rural communities hit by the opening of trade markets, or mining communities struck by the depletion of natural resources. The patterns and consequences experienced by these small cities are different from those in large cities, so planning interventions should be different. Second, future studies need to add more empirical investigations into how planning policies have been successfully performed. Many studies have argued for the reuse of vacant or abandoned properties, but performance evaluation of reuse approaches is still quite limited. More broadly, performance evaluations of smart declining strategies such as land banking, land-use adjustment, reduction of infrastructure, or relocation assistance are necessary before expanding and adopting such tools routinely. Third, shrinking cities research has gone full circle from an emphasis on green infrastructure to policy, and back to an emphasis on green infrastructure. Although re-greening is important and can effectively counter the decline, future studies could be more creative and take different research approaches. For example, researchers may develop processes for finding areas that need the most help using tools such as suitability analyses, or by reclassifying spatial patterns for vacancy using machine learning. Researchers should also put more effort into techniques for projecting future decrease trends connected to shrinkage, based on recent trends. All these future study proposals would need increased collaboration among shrinkage researchers, who are not presently well linked, in order to be effective.

Biographies

Jung-Eun Lee

Jung-Eun Lee is a graduate student in the Department of Architectural and Urban Systems Engineering at Ewha Womans University in South Korea. Her research interests include urban planning, urban shrinkage, urban regeneration, and spatial analytics. She is focuses particularly on the causes and countermeasures of the shrinking cities.

Yunmi Park

Yunmi Park holds a Ph.D. in Urban and Regional Science from Texas A&M University, and is a certified planner in both Korea and the United States (AICP), with professional planning experience in South Korea. Prior to joining Seoul National University, she worked at Auburn University and Ewha Womans University. Her areas of expertise and research interests include smart city planning and policy, urban shrinkage and regeneration, and spatial analytics.

Galen D. Newman

Galen D. Newman is an associate professor in the Department of Landscape Architecture and Urban Planning at Texas A&M University. His research interests include urban regeneration, land use science, spatial analytics, urban resilience, and community/urban-scaled design. His research focuses on the integration of urban regeneration and urban flood resilience.